Impact of Dietary Shifts on Gut Microbiome Dynamics

Multivariate Insights Using R

R for Bio Data Analysis

Group 16: Eric Torres, Lucia de Lamadrid, Konstantina Gkopi, Elena Iriondo and Jorge Santiago

2024-12-03

Introduction

Metagenomics study focusing on an animal model of the human gut ecosystem obtained by transplanting fresh human fecal microbial communities into germ-free C57BL/6J mice.
Measuring the effects of diet (Western or Low-Fat Plant Polysaccharide-rich), temporal and spatial composition of the microbiota, colonisation history…

Our aim:

To study the relationship between the composition of the gut microbiota and factors such as diet and colonisation history.

Data Loading

MicrobiomeWithMetadata.csv (Main dataset):
Contains the relative abundance of OTUs analyzed per observation, along with the conditions of each sample (e.g., diet, sex, etc.).

library(tidyverse) 
# Read the data
metadata_df <- read_tsv(file = str_c('../data/01_data_metadata.tsv'))

# Check the structure of the data
head(metadata_df)

# A tibble: 6 × 6,701
   Diet Source Donor CollectionMet   Sex     OTU0     OTU1     OTU2     OTU3
  <dbl>  <dbl> <dbl>         <dbl> <dbl>    <dbl>    <dbl>    <dbl>    <dbl>
1     0      0     0             0     0 1.56e-11 4.72e-11 1.23e-11 4.52e-11
2     0      1     0             0     0 2.36e-11 9.53e-11 3.33e-11 2.67e-11
3     0      2     0             1     0 6.77e-11 3.68e-11 8.02e-11 5.49e-11
4     0      2     0             0     0 5.52e-11 9.89e-11 4.58e-11 3.54e-11
5     0      3     0             0     0 5.24e-11 6.34e-11 2.35e-11 7.47e-11
6     0      4     0             1     0 7.67e-11 7.22e-11 5.41e-11 1.20e-11
# ℹ 6,692 more variables: OTU4 <dbl>, OTU5 <dbl>, OTU6 <dbl>, OTU7 <dbl>,
#   OTU8 <dbl>, OTU9 <dbl>, OTU10 <dbl>, OTU11 <dbl>, OTU12 <dbl>, OTU13 <dbl>,
#   OTU14 <dbl>, OTU15 <dbl>, OTU16 <dbl>, OTU17 <dbl>, OTU18 <dbl>,
#   OTU19 <dbl>, OTU20 <dbl>, OTU21 <dbl>, OTU22 <dbl>, OTU23 <dbl>,
#   OTU24 <dbl>, OTU25 <dbl>, OTU26 <dbl>, OTU27 <dbl>, OTU28 <dbl>,
#   OTU29 <dbl>, OTU30 <dbl>, OTU31 <dbl>, OTU32 <dbl>, OTU33 <dbl>,
#   OTU34 <dbl>, OTU35 <dbl>, OTU36 <dbl>, OTU37 <dbl>, OTU38 <dbl>, …

MicrobiomeMetadataDictionary.csv (Complementary):
Provides the equivalence between coded numerical values and their corresponding sample conditions.
MicrobiomeOTUtaxonomy.csv (Complementary):
Contains the taxonomic classification of each OTU analyzed.
Source:
Explore Microbiome Dataset
Curated version of the microbiome dataset from Turnbaugh et al. (original publication available in accompanying files).

Materials and Methods

Data Tidying and Filtering

Added a SampleID column to uniquely identify each sample + Transformed the dataset from wide to long format for easier analysis.

# Creation and relocation of SampleID
metadata_df <- metadata_df |>
  mutate(SampleID = row_number()) |>  # Create SampleID from the first column
  relocate(SampleID, 
           .before = everything())  # Move SampleID to the first position

metadata_df_long <- metadata_df |> 
  pivot_longer(
    cols = starts_with("OTU"), 
    names_to = "OTU", 
    values_to = "rel_abundance"
  )
head(metadata_df_long)

# A tibble: 6 × 8
  SampleID  Diet Source Donor CollectionMet   Sex OTU   rel_abundance
     <int> <dbl>  <dbl> <dbl>         <dbl> <dbl> <chr>         <dbl>
1        1     0      0     0             0     0 OTU0       1.56e-11
2        1     0      0     0             0     0 OTU1       4.72e-11
3        1     0      0     0             0     0 OTU2       1.23e-11
4        1     0      0     0             0     0 OTU3       4.52e-11
5        1     0      0     0             0     0 OTU4       2.72e-11
6        1     0      0     0             0     0 OTU5       2.71e-11

Filtering Low-Abundance OTUs: Retained OTUs contributing up to 95% of cumulative abundance. Additionally, removed OTUs with extremely low relative abundances.
Replaced the numeric codes with descriptive labels.

# Calculate cumulative contribution
cumulative_otus <- metadata_df_long |>
  group_by(OTU) |>
  summarize(mean_abundance = mean(rel_abundance)) |>
  arrange(desc(mean_abundance)) |>
  mutate(cumulative_abundance = cumsum(mean_abundance) / sum(mean_abundance))
# Filter OTUs contributing to 95% cumulative abundance
otus_to_keep <- cumulative_otus |>
  filter(cumulative_abundance <= 0.95) |>
  pull(OTU)
# Filter the metadata to retain only these OTUs
filtered_metadata <- metadata_df_long |>
  filter(OTU %in% otus_to_keep)
dim(filtered_metadata)

[1] 264600      8

# Set the stricter abundance threshold
abundance_threshold <- 1e-6
# Apply the threshold to filter OTUs
filtered_metadata_stricter <- filtered_metadata |>
  filter(rel_abundance >= abundance_threshold)
dim(filtered_metadata_stricter)

[1] 65938     8

filtered_metadata_stricter_label <- filtered_metadata_stricter |> 
  mutate(Diet = case_when(Diet == 0 ~ "LFPP",
                          Diet == 1 ~ "Western",
                          Diet == 2 ~ "CARBR",
                          Diet == 3 ~ "FATR",
                          Diet == 4 ~ "Suckling",
                          Diet == 5 ~ "Human")) |> 
  mutate(Source = case_when(Source == 0 ~ "Cecum1",
                          Source == 1 ~ "Cecum2", 
                          Source == 2 ~ "Colon1", 
                          Source == 3 ~ "Colon2", 
                          Source == 4 ~ "Feces",
                          Source == 5 ~ "SI1",
                          Source == 6 ~ "SI13", 
                          Source == 7 ~ "SI15", 
                          Source == 8 ~ "SI2", 
                          Source == 9 ~ "SI5",
                          Source == 10 ~ "SI9", 
                          Source == 11 ~ "Stomach", 
                          Source == 12 ~ "Cecum")) |> 
  mutate(Donor = case_when(Donor == 0 ~ "HMouseLFPP",
                          Donor == 1 ~ "CONVR", 
                          Donor == 2 ~ "Human", 
                          Donor == 3 ~ "Fresh", 
                          Donor == 4 ~ "Frozen",
                          Donor == 5 ~ "HMouseWestern", 
                          Donor == 6 ~ "CONVD")) |> 
  mutate(CollectionMet = case_when(CollectionMet == 0 ~ "Contents",
                                   CollectionMet == 1 ~ "Scraping")) |> 
  mutate(Sex = case_when(Sex == 0 ~ "Male",
                         Sex == 1 ~ "Female"))

:::

Now, our data is tidy!

clean_df <- read_tsv('../data/02_metadata_long_filtered_label.tsv')
head(clean_df)

# A tibble: 6 × 8
  SampleID Diet  Source Donor      CollectionMet Sex   OTU   rel_abundance
     <dbl> <chr> <chr>  <chr>      <chr>         <chr> <chr>         <dbl>
1        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU6       3.31e-11
2        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU7       5.08e-11
3        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU9       2.57e- 3
4        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU41      7.95e-11
5        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU58      2.53e-11
6        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU77      1.28e- 3

and ready to augment it…

Our data is tidy… and ready to be augmented!

We will use the OTUs taxonomy file to add columns with the names of phylum and class for each OTU, using left_join.

# A tibble: 6 × 8
  SampleID Diet  Source Donor      CollectionMet Sex   OTU   rel_abundance
     <dbl> <chr> <chr>  <chr>      <chr>         <chr> <chr>         <dbl>
1        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU6       3.31e-11
2        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU7       5.08e-11
3        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU9       2.57e- 3
4        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU41      7.95e-11
5        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU58      2.53e-11
6        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU77      1.28e- 3

:::
:::
:::

clean_df_taxonomy <- clean_df |>  
  left_join(otu_df_modified, 
            join_by(OTU == OTU.ID)) |> 
  relocate(Phylum, Class, .after = OTU) 

head(clean_df_taxonomy)

# A tibble: 6 × 10
  SampleID Diet  Source Donor      CollectionMet Sex   OTU   Phylum     Class   
     <dbl> <chr> <chr>  <chr>      <chr>         <chr> <chr> <chr>      <chr>   
1        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU6  Firmicutes Bacilli 
2        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU7  Firmicutes Clostri…
3        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU9  Firmicutes Clostri…
4        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU41 Firmicutes Bacilli 
5        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU58 Firmicutes Clostri…
6        1 LFPP  Cecum1 HMouseLFPP Contents      Male  OTU77 Firmicutes Clostri…
# ℹ 1 more variable: rel_abundance <dbl>

Results and Discussion

Microbiota composition in terms of phyla in different:

sources and diet types

diet and donor combination

05 Principal Component Analysis on Phylum-Level Aggregated Microbiome Data

# A tibble: 6 × 9
  SampleID  Diet Actinobacteria Bacteroidetes Firmicutes Proteobacteria      TM7
     <dbl> <dbl>          <dbl>         <dbl>      <dbl>          <dbl>    <dbl>
1      339     1        0.00316        0.463       0.470        0.00671 7.70e-11
2      340     1        0.0180         0.106       0.811        0.00138 3.72e-11
3      341     1        0.0121         0.0443      0.873        0.00792 5.68e-11
4      342     1        0.00531        0.145       0.786        0.0135  7.89e-12
5      343     1        0.00801        0.607       0.300        0.0300  3.90e-11
6      344     1        0.631          0.199       0.141        0.00467 3.98e-11
# ℹ 2 more variables: Unclassified <dbl>, Verrucomicrobia <dbl>

Principal Component Analysis on Phylum-Level Aggregated Microbiome Data

Analysis of Microbiome Clusters by Donor Groups Using Hierarchical Clustering

We compute a distance matrix using the Euclidean distance to measure similarity between microbiome samples. The clustering is performed using Ward’s method, which minimizes variance within clusters.

# Compute Euclidean distance matrix
dist_matrix <- otu_data_scaled |>
  dist()

# Perform hierarchical clustering
hclust_result <- hclust(dist_matrix, method = "ward.D2")

# Cut dendrogram into 3 clusters
cluster_labels <- cutree(hclust_result, k = 3) |>
  as_tibble() |>
  rename(Cluster = value)

Chi-Squared Test

# Perform chi-squared test
chi2_result <- chisq.test(donor_cluster_table)
chi2_result


    Pearson's Chi-squared test

data:  donor_cluster_table
X-squared = 659.91, df = 8, p-value < 2.2e-16

Cluster 1 is dominated by HMouseLFPP (55.5%) with notable contributions from Frozen (17.8%) and Fresh (18.7%), reflecting plant-rich diets and preserved samples.
Cluster 2 includes mostly Fresh (55.1%) and HMouseLFPP (26.5%), indicating a mix of human-derived and dietary influences.
Cluster 3 is almost entirely CONVR (95%), representing natural microbiota from control mice.
The chi-squared test confirms significant associations between donor origins and clusters, highlighting the influence of donors on microbiota composition.

Biodiversity and diet

Shannon diversity index

Number of species living in a habitat (richness)

Relative abundance (evenness).

\[ H' = -\sum_{i=1}^R p_i \ln p_i \\ p_i \text{ is the relative abundance of OTU}_i\\ R \text{ is the total number of OTUs} \]

Biodiversity in the microbiota of first-generation humanized mice was found to differ significantly across different diets.

Conclusion

The “Obesity-inducing” diet influences the Firmicutes-Bacteroidetes ratio

PCA shows how diet shapes microbial composition, as well as the relationship between different phyla.
Clustering shines light on how the microbiota donor structures the data
The Western diet favours a more biodiverse gut ecosystem